[Paper Review] First Results of $ u_e$ Appearance Analysis and Electron Neutrino Identification at NOvA

J. M. Bian|arXiv (Cornell University)|Oct 19, 2015

Neutrino Physics Research5 references9 citations

TL;DR

This paper presents the first evidence of νµ → νe oscillation in the NOvA experiment using 2.74×10²⁰ protons-on-target, achieved through a novel electron neutrino identification (EID) algorithm based on 3D shower-energy profile likelihoods and an artificial neural network. The analysis yields a 3.3σ significance for νe appearance, favoring the normal neutrino mass hierarchy and δCP ≈ 3π/2, with strong constraints on sin²2θ₁₃ consistent with reactor measurements.

ABSTRACT

NOvA is a long-baseline accelerator-based neutrino oscillation experiment that is optimized for $ u_\mu o u_e$ measurements. It uses the upgraded NuMI beam from Fermilab and measures electron-neutrino appearance and muon-neutrino disappearance at its Far Detector in Ash River, Minnesota. The $ u_e$ appearance analysis at NOvA aims to resolve the neutrino mass hierarchy problem and to constrain the CP-violating phase. The first data set of $2.74 imes10^{20}$ protons on target (POT) equivalent exposure taken by NOvA has been analyzed. The first measurement of electron-neutrino appearance in NOvA provides solid evidence of $ u_\mu o u_e$ oscillation with the NuMI beam line. Electron-neutrino identification is the key ingredient for the $ u_e$ appearance analysis. The electron-identification algorithm used to produce the primary results presented here compares 3-D shower-energy profiles with Monte Carlo prototypes to construct likelihoods for each particle hypothesis. Particle likelihoods, among other event-topology variables, are used as inputs to an Artificial Neural Network for the final electron-neutrino identification. The design and implementation of this algorithm is also presented.

Motivation & Objective

• To measure νµ → νe oscillation in a long-baseline neutrino experiment to resolve the neutrino mass hierarchy and constrain the CP-violating phase.
• To develop and validate a high-precision electron neutrino identification (EID) algorithm for distinguishing νe CC events from backgrounds like π⁰ and muons.
• To provide the first νe appearance measurement using the NuMI beam, leveraging off-axis beam focusing and high granularity to enhance signal sensitivity.
• To reduce systematic uncertainties through data-driven extrapolation from the Near Detector to the Far Detector.

Proposed method

• The EID algorithm, LID, uses 3D shower-energy profiles to compute likelihoods for electron, photon, muon, and pion hypotheses by comparing measured dE/dx to Monte Carlo simulations plane-by-plane and cell-by-cell.
• A multi-layer perceptron neural network combines particle likelihoods and event topology variables to produce a final electron-neutrino identification score.
• Shower clustering is performed using a modified Hough transform to reconstruct interaction vertices and prong directions from hit clusters in time-coincident slices.
• Cell energy deconvolution is applied to overlapping showers using a lateral energy profile model with λ = 3.05 cm, assuming exponential attenuation from the shower core.
• Event selection applies cuts on total energy (1.5–2.7 GeV), transverse momentum (PT/P > 0.8), and PID score (LID > 0.95, LEM > 0.8) to enrich νe CC candidates.
• Backgrounds are predicted using Near Detector data to correct Monte Carlo normalization, with systematic uncertainties evaluated via variations in calibration, response non-linearity, and interaction models.

Experimental results

Research questions

• RQ1 What is the significance of νµ → νe oscillation in the NOvA experiment with the first 2.74×10²⁰ POT of data?
• RQ2 Can the LID electron identification algorithm effectively suppress backgrounds from π⁰, muons, and cosmic rays in a high-granularity liquid scintillator detector?
• RQ3 What constraints can be placed on the neutrino mixing angle sin²2θ₁₃ and CP-violating phase δCP from the first νe appearance measurement?
• RQ4 Does the observed νe candidate excess favor the normal or inverted neutrino mass hierarchy?

Key findings

• The νe appearance analysis observes 6 events with LID > 0.95 and 11 with LEM > 0.8, with 6 events selected by both, yielding a 3.3σ significance for νµ → νe oscillation.
• The LEM-based analysis yields a higher significance of 5.4σ, with 12% of pseudo-experiments in the normal hierarchy, δCP = 3π/2 scenario having a worse χ² than observed.
• The 68% confidence level region for the normal mass hierarchy is compatible with the world average reactor result sin²2θ₁₃ = 0.086 ± 0.05 across all δCP values.
• The analysis disfavors the inverted mass hierarchy in the range 0 < δCP < 0.65π at 90% C.L. for sin²θ₂₃ = 0.5, with both LID and LEM favoring δCP near 3π/2.
• The background prediction is 0.94 ± 0.10 (syst.) events for LID and 1.00 ± 0.12 (syst.) for LEM, dominated by beam νe CC (49–46%) and neutral current (38–40%) processes.
• The signal prediction reaches 5.62 ± 0.99 events under the normal hierarchy, δCP = 3π/2, θ₂₃ = π/4 assumption, consistent with data and MC validation.